Impact of different convective cloud schemes on the

simulation of the tropical seasonal cycle in a coupled

ocean-atmosphere model.

Pascale Braconnot, Frederic Hourdin, Sandrine Bony, Jean-Louis Dufresne,

Jean-Yves Grandpeix, Olivier Marti

To cite this version:

Pascale Braconnot, Frederic Hourdin, Sandrine Bony, Jean-Louis Dufresne, Jean-Yves Grand-peix, et al.. Impact of different convective cloud schemes on the simulation of the tropicalseasonal cycle in a coupled ocean-atmosphere model.. Climate Dynamics, Springer Verlag,2007, 29, pp.501-520. <10.1007/s00382-007-0244-y>. <hal-00184738>

HAL Id: hal-00184738

https://hal.archives-ouvertes.fr/hal-00184738

Submitted on 31 Oct 2007

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinee au depot et a la diffusion de documentsscientifiques de niveau recherche, publies ou non,emanant des etablissements d’enseignement et derecherche francais ou etrangers, des laboratoirespublics ou prives.

5/29/2006 1

Impact of different convective cloud schemes on the simulation of the tropical seasonal cycle with a coupled ocean-atmosphere model P. Braconnot1*, F. Hourdin2, S. Bony2, J.-L. Dufresne2, J.-Y. Grandpeix2, and O. Marti1

1. IPSL/LSCE, unité mixte CEA-CNRS-UVSQ, Bât.712, Orme des Merisiers, 91191 Gif-sur-Yvette Cedex. * Corresponding author. 2. IPSL/LMD, case 99, 4 place Jussieu, 75252 Paris cedex 01.

Abstract.

The simulation of the mean seasonal cycle of sea surface temperature (SST) remains a

challenge for coupled ocean-atmosphere general circulation models (OAGCMs). Here we

investigate how the numerical representation of clouds and convection affects the simulation of the

seasonal variations of SST. For this purpose, we analyze the simulations of two versions of the

same OAGCM differing only by their convective cloud schemes. Most of the differences between

the two simulations in the mean atmospheric temperature and precipitation fields reflect differences

found in atmosphere-only simulations. They affect the ocean interior down to 1000 m.

Substantial differences are found between the two coupled simulations in the seasonal march of

the Intertropical Convergence Zone in the eastern part of the Pacific and Atlantic basins, where the

equatorial upwelling develops. The results show that the convergence of humidity between ocean

and land during the American and African boreal summer monsoons plays a key role in maintaining

a cross equatorial flow and a strong windstress along the equator. An other important factor is the

presence of low level clouds in subsidence regions over cold waters south of the equator, which

favors the maintenance of a cross equatorial SST gradient and of a surface wind convergence north

of the equator. The results suggest that the inhibition of convection by downdraughts, as well as

radiative feedbacks between clouds and SST, have a key role on the relative distribution of large-

scale convection between ocean and land, and thereby the dynamics of the equatorial upwelling and

the SST seasonal cycle in the Pacific and Atlantic oceans.

5/29/2006 2

1. Introduction

The representation of clouds and atmospheric convection has been emphasized as a critical

issue for the simulation of the current climate in various coupled ocean atmosphere models (IPCC

2001). In particular coupled models have a tendency to produce a warm bias in the eastern Pacific

and Atlantic oceans where the equatorial upwelling develops from the eastern border (Latif et al.

2001; Davey et al. 2002). The representation of the seasonal cycle of Sea surface temperature

(SST), such as the representation of the seasonal evolution of the equatorial upwelling in the eastern

Pacific remains a challenge for coupled simulations. It is an example of a coupled ocean-

atmosphere variability potentially strongly affected by the numerical representation of clouds and

convection. Indeed, the existence of the cold tongue along the equator is an annual response of the

climate system to the semi-annual forcing of incoming solar radiation at the top of the atmosphere

in this region. It results from complex atmosphere-ocean interaction (Wang 1994). Two modes have

been identified so far that shape the seasonal evolution of SST in this region (Wang 1994; Nigam

and Chao 1996). The first one is related to the monsoon circulation that converges in central

America. This mode represents the tropical part of a global response of the atmosphere-land-ocean

system to the differential insolation between the northern and southern hemisphere (Wang 1994).

The second mode results from large-scale air-sea interactions, and involves the relationship between

SST gradients and wind, the latent heat flux and the presence of a stratocumulus deck in the

subsiding regions of the east Pacific. Several studies have shown that the interactions between low-

level moisture , stratocumulus deck and cloud radiative properties contribute to the maintenance of

the cold tongue (Ma et al. 1996; Yu and Mechoso 1999; Gordon et al. 2000).

In that vein, we investigate how atmospheric convection and clouds impact the large-scale

atmospheric and oceanic circulation in the tropical regions, using two versions of a coupled ocean-

atmosphere models differing only by the parameterisation of convective clouds. The corresponding

simulations show substantial differences in the tropical circulation. In particular the seasonal

evolution of the Intertropical Convergence Zone (ITCZ) in the eastern part of the tropical Pacific

5/29/2006 3

and Atlantic is more confined towards the equator in one of the simulations and the strength of the

equatorial upwelling is damped. The comparison of these two simulations allows us to check how

this is related to the land-sea contrast and to the simulation of cloud cover in the eastern Pacific. The

objective of the paper is thus to highlight differences in the adjustment of the ocean-atmosphere

system to convective cloud schemes, and to show how the interaction between clouds, convection

and the large-scale circulation affects the characteristics of the mean climate and of the mean

seasonal cycle in tropics.

We use the last version of the IPSL coupled ocean-atmosphere model (Marti et al. 2005). Its

atmospheric component, LMDZ, has been recently improved as described by Hourdin et al. (2006).

The Tiedke (1989)’s convection scheme, originally implemented in the LMDz atmospheric model

by Li (1999), was carefully tuned so that it was possible to run a millennium time scale coupled

simulation with a previous version of the IPSL coupled model, for which sea-ice was prescribed to

climatological values and the land surface was represented using a simple bucket model (Li and

Conil 2003). Replacement of the Tiedke (1989)’s scheme by the Emanuel (1993)’s scheme greatly

improved the large scale distribution of tropical precipitation in simulations with the atmospheric

model. In particular, the Inter Tropical Convergence zone (ITCZ) is less confined along the equator

with the Emanuel ‘s scheme, and precipitations are better reproduced over the Amazonian basin

(Hourdin et al. 2006). The rate of warming and moistening of the atmospheric column by

convection shows substantial differences between the two schemes, particularly over ocean. This

difference is particularly marked over ocean, where the convective downdraughts of the Emanuel's

scheme decrease the surface layer moist static energy much more that those of the Tiedtke's scheme.

Analyses of atmosphere-alone simulations with the two convection schemes suggest that these

differences are at the origin of substantial differences in the simulation of the land-sea contrast in

convection and of the large scale tropical atmospheric circulation (Hourdin et al. 2006). Our

purpose here is to analyse how these differences impact the simulated climate when the atmosphere

is coupled to the ocean.

5/29/2006 4

The remainder of the manuscript is organized as follows. Section 2 presents a short description

of the coupled model, with emphasis on the convection and cloud schemes, and on the major

characteristics of the simulations. Differences in the tropical circulation simulated with the two

convection schemes are analysed in section 3. Section 4 discusses differences in the timing and in

the amplitude of the tropical seasonal cycle. The mechanisms responsible for the differences in the

seasonal march of the Intertropical Convergence Zone (ITCZ) in the eastern equatorial Pacific and

Atlantic oceans are discussed in section 5. Comparison with atmosphere alone simulations helps us

to understand the role of the interactions between convection and the large scale circulation in the

tropics. The conclusion is provided in section 6.

2. Model and simulations

Model description

Simulations were performed with the last version of the coupled model developed at Institut

Pierre Simon Laplace (IPSL, Paris Marti et al. 2005). IPSL_CM4 couples the grid point

atmospheric general circulation model LMDz (Hourdin et al. 2006) developed at Laboratoire de

Météorologie Dynamique (LMD, France) and the oceanic general circulation model ORCA (Madec

et al. 1998) developed at Laboratoire d'Océanographie Dynamique et de Climatologie (LOCEAN,

France). On the continent, the land surface scheme ORCHIDEE (Krinner et al. 2005) is coupled to

the atmospheric model. Only the thermodynamic component of ORCHIDEE is active in the

simulations presented here. The evolution of the carbon cycle and the dynamics of the vegetation

are not considered. The closure of the water budget with the ocean is achieved thanks to a river

routing scheme implemented in the land surface model. A sea-ice model (Fichefet and Maqueda

1997), which computes ice thermodynamics and dynamics, is included in the ocean model. The

ocean and atmospheric models exchange surface temperature, sea-ice cover, momentum, heat and

fresh water fluxes once a day, using the OASIS coupler (Terray et al. 1995) developed at

CERFACS (France). None of these fluxes is corrected.

5/29/2006 5

Compared to previous versions of the IPSL model (Braconnot et al. 2000; Dufresne et al.

2002), the coupling scheme was revisited to ensure both global and local conservations of

momentum, heat and fresh water fluxes at the air-sea interface. Following some of the ideas

implemented in previous model versions (Dufresne and Grandpeix 1996; Braconnot 1998), an

interface model was implemented in the atmospheric model. Each atmospheric grid box accounts

for up to four different subsurface types: ocean, land, sea-ice, and land ice. Surfaces fluxes are

computed according to the surface characteristics (roughness length, stability etc..). The

interpolation scheme conserves the heat and water budget both globally and locally (Marti et al.

2005).

In its present configuration, the model is run at medium resolution. The atmospheric grid is

regular, with a resolution of 3.75° in longitude, 2.5° in latitude, and 19 vertical levels. The ocean

model grid has approximately 2 degrees resolution (0.5 degrees near the equator) with 184 points in

longitude, 142 points in latitude and 31 1evels in the ocean.

Convection and cloud scheme

The coupled model is run either with Emanuel’s or Tiedke’s convection scheme. These schemes

both belong to the category of "mass flux schemes". They parameterize the convective mass fluxes

as well as the induced motions in the environmental air. In the Tiedtke's (1989) scheme, only one

convective cloud is considered, comprising one single saturated updraught. Entrainment and

detrainment between the cloud and the environment can take place at any level between the free

convection level and the free sinking level. There is also one single downdraught extending from

the free sinking level to the cloud base. The mass flux at the top of the downdraught is a constant

fraction (here 0.3) of the convective mass flux at cloud base. This downdraught is assumed to be

saturated and is kept at saturation by evaporating precipitation. The version of the Tiedtke's scheme

used here is close to its original formulation and uses a closure in moisture convergence. Triggering

is a function of the buoyancy of lifted parcels at the first grid level above condensation level.

5/29/2006 6

In the Emanuel’s scheme, the backbone of the convective systems is regions of adiabatic ascent

originating from some low-level layer and ending at their level of neutral buoyancy. Shedding from

these adiabatic ascents yields, at each level, a set of draughts which are mixtures of adiabatic ascent

air (from which some precipitation is removed) and environmental air. These mixed draughts move

adiabatically up or down to levels where, after further removal of precipitation and evaporation of

cloud water, they are at rest at their new levels of neutral buoyancy. In addition to those buoyancy-

sorted saturated draughts, unsaturated downdraughts are parameterized as a single entraining plume

of constant fractional area (here 1% of the grid cell) driven by the evaporation of precipitation. The

version of the Emanuel's scheme used here is very similar to Emanuel (1993). Closure and

triggering take into account both tropospheric instability and convective inhibition. Our version

differs from the original one in the removal of most explicit grid dependencies (e.g. the lifting

condensation level varies continuously and not from grid level to grid level), which yields a

smoother variation of convection intensity with time and a weaker dependence on vertical

resolution.

The cloud scheme of the model is a statistical cloud scheme that describes the subgrid-scale

distribution of total water by a generalized log-normal distribution, whose variance and skewness

are diagnosed, and that uses zero as the lower bound of the distribution. Non-convective clouds are

predicted by imposing that the statistical moments of the total water distribution are a function of

the vertical profile of total water and of saturated specific humidity (Hourdin et al. 2006). The non-

convective clouds are exactly the same whatever the convection scheme used. Clouds associated

with cumulus convection are predicted differently depending on the convection scheme used in the

GCM. When using the Emanuel’s scheme, the cloud scheme is coupled to the convection scheme

by diagnosing the variance and the skewness of the total water distribution from the vertical profile

of in-cloud water content predicted by the convection scheme and from the degree of saturation of

the large-scale environment (Bony and Emanuel 2001). In the case of the Tiedke’s scheme, an

homogeneous cloud fraction is assumed between cloud base and cloud top, that is a function of the

5/29/2006 7

vertically-integrated moistening tendency predicted by the Tiedtke's convection scheme (Hourdin et

al. 2006).

Cloud microphysical properties are computed as described by Bony and Emanuel (2001, Table

2 for water clouds and case "ICE-OPT" of Table 3 for ice clouds): temperature thresholds (-15°C

and 0°C) are used to partition the cloud condensate into liquid and frozen cloud water mixing ratios;

cloud optical thickness is computed by using an effective radius of cloud particles set to a constant

value for liquid water clouds (12 μm in the simulations presented here), and decreasing with

decreasing temperature (from 60 to 3.5 μm) for ice clouds. The vertical overlap of cloud layers is

assumed to be maximum-random. A more complete description of these schemes can be found in

Hourdin et al. (2006).

Simulations

We consider two coupled simulations run with these two schemes. Both schemes are equivalent

from a global energetic point of view. This means that the coupled model is equilibrated and that

long term integrations are possible without using any flux correction at the air-sea interface.

The simulation with the Emanuel’s scheme, referred to as KE, corresponds to the reference

version of the IPSL model that has been used for the production of IPCC AR4 scenarios (Dufresne

et al. 2005; Swingedouw et al. 2006). This simulation will then be considered as the reference. It is

a 500 year long simulation. Over this period, the model drift in global mean surface temperature is

less than 0.02 °C/century. The simulation is stable with a global mean 2m air temperature of 14°C

that presents low frequency variability of about 0.1 to 0.2°C around this mean. In the following, we

will consider a mean seasonal cycle computed from the last 200 years of the simulation.

The simulation with the Tiedke scheme (TI) was run for 300 years. The initial state and the spin

up strategy is the same as for KE. The model drift is negligible, but larger than for the KE

simulation (0.05°C/century). The global mean temperature is warmer (15.7 °C) and the low

frequency variability smaller. We will consider the last 150 years of the simulation in the remainder

5/29/2006 8

of this manuscript. When needed, the results will be compared with the atmosphere-alone

simulations presented by Hourdin et al. (2006).

Simulated climate: global characteristics.

Figure 1 shows the difference between the simulated annual mean sea surface temperature

(SST) and the Levitus (1982)’s climatology for the two simulations. The large-scale patterns of the

SST field are successfully captured by the two model versions in most places.

The major default of the reference simulation (KE) is the cold bias in the mid latitudes both in

the Atlantic and Pacific oceans (Figure 1a). Locally, this bias reaches 6 °C in the North Atlantic,

which results from the conjunction of a southward shift of the north Atlantic drift and the advection

of cold water from the Gin and the Labrador seas. It has been attributed to a lack of deep water

formation in the Labrador Sea (Swingedouw et al. 2005), caused by an overestimation of the

atmospheric fresh water budget over the North Atlantic, the Nordic seas and the southern part of

Greenland. This excessive fresh water input creates a thin halocline at the surface, which reduces

the local buoyancy of surface waters, and thus ocean convection. Note that the equatorward shift of

the mid-latitude stormtracks is already present in atmosphere alone simulations. It is amplified

when the atmospheric model is coupled to the ocean and leads to the mid latitude cold SSTs in both

hemisphere. The shift reduces when increasing horizontal resolution (not shown).

This cold bias is not as large in the TI simulation (Figure 1b). On the contrary, surface waters

are too warm along the margin of sea-ice in the southern hemisphere and in the northern part of the

Pacific ocean. The root mean square differences (rms) between the simulated map and Reynold’s

(1988) observations suggest that TI better reproduces global SST with a global rms of 1.4 °C

against 1.7 °C for KE. When the rms is computed over the tropical regions (30°S-30°N), the rms is

1.5 °C for TI and 1.2 °C for KE. This suggests that model biases are primarily located in mid-

latitudes in KE, whereas they occur at most latitudes in TI.

The major differences between the two coupled simulations and observations follow the

differences highlighted in atmosphere-alone simulations (Hourdin et al. 2006). For example, figure

5/29/2006 9

2a shows that in KE the zonal mean temperature presents a cold bias of up to 8°C at about 200hPa.

This pattern was attributed to too large moisture content at these levels and is amplified by the

coupling with the ocean. Differences in the atmosphere have their counterpart in the ocean. The KE

simulation presents a slight cold bias on the first 400 m of the ocean around 30°N and 40°S,

consistent with the cold atmospheric bias (Figure 2a). Elsewhere in the ocean, the departure from

climatology doesn’t exceed 2°C, except at high latitudes, where the warm bias at depth reflects the

lack of deep water formation.

Figure 2b also reveals that in middle and high latitudes, the TI simulation is associated with

even colder conditions in the upper troposphere and lower stratosphere and warmer conditions in

the lower and middle troposphere than the KE simulation. The TI simulation is also characterized

by warmer water in the first 1000 m of the ocean with maximum value in the two regions of

intermediate water formation noted above. It is less satisfactory at depth were 1°C warmer than

observed water is found down to 1000 m.

The equator-to-pole differences in the tri-dimensional structure of temperature in the ocean and

atmosphere is associated with large-scale differences in the dynamics of the two fluids. In

particular, the northward heat transport is larger in KE, both in the atmosphere and in the ocean.

These differences reach about 1PW at 20°N. For this aspect, KE is in better agreement with

Trenberth and Solomon (1994) estimate of the heat transport from satellite observation and

reanalysis.

3. Large scale features in the tropics

Large scale features of the simulated climate compared to observations.

The model performs well in the tropics. In KE, errors are less than 0.5°C over most of the

basins (figure 1a). The exception is at the eastern side of the equatorial Pacific and Atlantic were the

warming indicates that the equatorial upwelling is not well located along the coast. This is a

classical bias of coupled model, especially with the relatively coarse resolution we are using here

(Latif et al. 2001). The equatorial section of annual mean temperature plotted as a function of depth

5/29/2006 10

(figure 3) shows that the vertical temperature gradient is more diffuse than in the observation.

However, the shape of the thermocline presents an east-west gradient in agreement with observation

both in the Pacific and Atlantic oceans. In the latter, warm surface waters accumulate in the first

ocean layers in the eastern part of the Gulf of Guinea and the upwelling is shifted to the west. The

surface ocean temperature as well as the east-west temperature gradient at depth are also

satisfactorily reproduced in KE in the Indian ocean (Figure 3b).

Figures 1 and 3c show that TI is slightly warmer than KE in the eastern Pacific and colder over

the warm pool. As a result, the East-West SST gradient is reduced by about 1.4 °C in TI compared

to KE across the Pacific. There is even more difference between the two simulations in the ocean

interior (Figure 3). For example, the 14°C isotherm depth is located about 100m deeper in the

eastern part of the Pacific Ocean, suggesting that the upwelling is not strong enough. In the Atlantic

TI produces a flat thermocline across the basin. This bias leads to a reversal of the temperature

gradient at the surface compared to observation, and the equatorial upwelling nearly vanishes. Also

the temperature gradients are reversed across the Indian ocean in the first 200 m, with warmer water

in the eastern part. This rapid comparison shows that the thermocline depth is on average deeper in

TI than in KE, which suggests that the local thermal inertia of the ocean is slightly higher in this

simulation.

The large scale characteristics of wind stress and precipitation (Figure 4 and 5) are consistent

with the temperature field. KE reproduces the broad features found in observations. In January, the

trade winds are well reproduced in the Pacific ocean except in the eastern part where they penetrate

too far south (Figure 4b, left). This structure is consistent with the lack of precipitation north of the

equator in this region and with the presence of a double ITCZ structure around 4°S (Figure 5b and

d, left). Along the coast of South America and South Africa the northward component of the

windstress is not strong enough (Figure 4b, left). The coastal upwelling is thus slightly damped and

the surface SST warmer than observed (Figure 1a). The windstress structure also is too zonal

between 4°S and the equator in the western Pacific in January (Figure 4b, left), consistent with the

zonal distribution of the southern Pacific convergence zone (SPCZ, Figure 5c and d, left). In July,

5/29/2006 11

the observed windstress is more intense in the southern hemisphere and up to 8°N in the eastern

Pacific and Atlantic Ocean (Figure 4a, right). This pattern is associated with the strengthening of the

northern hemisphere Hadley cell, and with the intensification of the monsoon flow from the ocean

towards the continent, where heavy precipitation occurs (Figure 5a, right). This return flow from the

ocean to the continents over America and Africa is reproduced in KE. However, windstress is not

strong enough along the eastern border in July (Figure 4b, right), consistent with the accumulation

of warm water in these regions and with the westward shift of the core of the upwelling (Figure 1a).

Figure 5d also shows that monsoon rainfall is underestimated over the southern border of North

America, Africa and India.

These drawbacks are more pronounced in TI. In January the trade winds are too strong in the

northern hemisphere and converge too far south (Figure 4c, left), which is consistent with the

southward shift of the ITCZ between the two simulations (Figure 5e, left) and with the too strong

precipitation over the maritime continent (Fig. 5c left). The maximum wind stress is also slightly

shifted from the middle of the basin to the west, and there is almost no extension of the eastward (or

the westerly) winds in the western Pacific (Fig. 4c, left). The zero line for both the zonal and

meridional components of the wind stress (not shown) is shifted to the south by about 5°degrees of

latitude. The Walker cell across the Indian Ocean is weaker in this simulation, which explains the

reversal of the temperature gradient in the ocean interior (Figure 3). The same weakening was

obtained in atmosphere-alone simulations. In July the return flow from the ocean to the continent

associated with the American and African monsoon is not strong enough (Fig. 4c, right). In the

Atlantic the Southeaststerlies do not penetrate far enough during summer in the northern

hemisphere. The wind stress is thus not strong enough along the equator and does not allow for the

development of the equatorial upwelling at the coast. It is also interesting to note that the ITCZ is

shifted to the south in TI compared to KE year round compared to KE in the Pacific ocean (Figure

5e). It is consistent with the too intense trade winds, their excessive extension in the southern

hemisphere in January, and their lack of northern extension in the eastern Pacific in July.

5/29/2006 12

The above discussion shows that representation of convective clouds in the model affects the

simulation of the tropical climate, and that the monsoon flow over the ocean, the July monsoon

precipitation over Central America and Africa, and the development of the equatorial upwelling in

the eastern Pacific and Atlantic oceans are connected. We analyze below how this connection

affects the seasonal cycle of SST in the eastern Pacific and Atlantic.

Mean seasonal cycle. Figure 6 shows the mean seasonal evolution of SST and precipitation for several regions in the

tropics. The phasing of the annual cycle of SST and precipitation is satisfactorily represented for

KE in most regions. However, as discussed above (comments of figure 1 and 5), precipitation is

underestimated in the West Pacific north of the equator from April to October. In the eastern part of

the basin, they are overestimated from January to June. This excessive precipitation is associated

with an excessive surface warming during the same period. This feature corresponds to the double

ITCZ structure that develops over a longer period and with a larger magnitude than in the

observations. A similar feature is present in the Atlantic ocean, where the SST is too warm from

November to June and where the precipitation is overestimated from January to June south of the

equator. The phasing of the annual mean cycle of SST and precipitation is less satisfactory in TI. In

particular the development of the double ITCZ and the associated warm surface conditions lasts

longer in this simulation and the magnitude is larger. The differences found in the Indian ocean in

SST (figure 1b and 3), are also associated with a misrepresentation of the double peak of SST in this

basin.

We display the geographic distribution of the seasonal cycle of the simulated SST in the tropics

and compare it with observations in figure 7. For simplicity, we characterize the seasonal cycle and

assess it against observations by considering its magnitude (quantified here as the difference

between the monthly maximum and the monthly minimum temperature estimated from a mean

seasonal cycle) and the correlation between the simulated and observed seasonal cycles of SST.

Results for the magnitude are plotted for each grid point as the ratio of the difference from

5/29/2006 13

observation over the observation. Figure 7a, b and c show that the two model versions tend to

underestimate the magnitude of the seasonal cycle in the eastern part of the Pacific and Atlantic

oceans, where SSTs are warmer than observed (Figure 1). This underestimate reaches up to 50% in

TI over large regions, whereas, it is around 20% in KE. On the other hand, the magnitude tends to

be overestimated in the western part of the basin. In the western Atlantic, we note an excessive

magnitude of the seasonal cycle in KE which is due to the westward shift of the equatorial

upwelling. The correlation between the observed and simulated seasonal cycles of the SST

characterizes the ability of the model to reproduce the phasing of the seasonal SST anomalies.

Figures 7 d and e show that it is better in KE in most places.

The evolution of the seasonal cycle along the equator averaged between 2°N and 2°S (Figure 8)

shows that both simulations produce a timing of the equatorial upwelling in phase with the

observations. However the westward shift of the upwelling core in the Pacific is more pronounced

in TI, as is the westward propagation with time (Figure 8). From experiments with an intermediate

coupled model, Fu and Wang (2001) suggested that the westward propagation of the cold tongue

was mainly due to the zonal advection of cold water by the oceanic circulation. This is the case in

TI where the zonal current is more intense west of 140°W due to the stronger trade winds in these

regions (Figure 4). In the Atlantic, the magnitude and the phase of the equatorial upwelling are

satisfactory in KE although its core is shifted to the west. As already noted, this equatorial

upwelling nearly vanishes in TI.

4. Tropical Oceans and seasonal march of the ITCZ

Seasonal march of the ITCZ in the eastern Pacific.

The analysis of the mean seasonal cycle shows that the convection and cloud schemes strongly

impact the representation of both the mean state and the seasonal characteristics of the climate in

the eastern part of the Pacific ocean. We now analyse this in more depth, considering first the region

extending from 110 °W to 90 °W and 8°S to 14 °S in the eastern Pacific (shaded in figure 6). This

region allows us to analyze the development of the equatorial upwelling and the connection

5/29/2006 14

between the equatorial SST and the location of the ITCZ. In this region SST has a well defined

seasonal cycle, with warmer water located south of the equator in boreal spring and extending to the

north of the equator from June to September when the equatorial and coastal upwelling intensifies

from 8°S to 2°N in boreal summer and autumn (Figure 9).

Several aspects of the wind and cloud fields have been identified as important players in this

feature. The upwelling along the coast of Peru and the equatorial upwelling in the east fully develop

when the wind is along the coast of South America and its strength intensifies to the north. At that

time, the zonal component of the windstress also intensifies along the equator towards the west,

which enhances the equatorial divergence and allows for the strengthening of the equatorial

upwelling further west. The onset of American summer monsoon on the continent at 15°N seems to

be an important factor in this intensification and in the position of the core of the equatorial

upwelling near the coast (Wang 1994). Several feedbacks, involving SST, clouds or wind and

evaporation enter then into play to reinforce the upwelling and maintain the ITCZ in a northern

position (e.g. Xie and Saito 2001).

We first consider KE. As already seen in figure 8, the seasonal evolution of the equatorial

upwelling is quite realistic in this simulation. In particular, the northward SST gradient between the

cold tongue and the warmer water to the north is present in KE from July to November (Figure 9).

This is consistent with the onset and the maintenance of the cross equatorial monsoon flow in this

simulation (Figure 4). Note however that the structure is slightly shifted towards the equator, and

that warm waters do not extend far enough to the north of 8°N, as suspected from the wind stress

and the lack of precipitation over the continent (Figure 5). During the time when the SST gradient

intensifies to the north (April-May), precipitation follows quite closely the region of maximum

surface warming in the northern hemisphere. In boreal winter and early spring, the maximum

precipitation is located as in the observations above warm waters around 4°S, peaking in March.

Figure 10 also shows that in KE the total cloud cover intensifies within the ITCZ and also south

of the equator over the cold waters from July to October. In the latter regions, air is subsiding and

clouds are limited to lower tropospheric levels. They reduce the incoming solar radiation at the

5/29/2006 15

ocean surface and locally favour the surface cooling. Therefore this version of the model produces

an annual variation of low level clouds consistent with a local strengthening of the meridional

circulation in the eastern Pacific as shown by Ma et al. (1996).

The two model versions exhibit a different behaviour in this region. Interestingly, the first part

of the northward shift of the warm waters from their equatorial position to 8°N in April-May is also

well reproduced in TI (Figure 9c, left), but as soon as the ITCZ intensifies to the north, the whole

structure shifts back to a more equatorward position and the SST gradient becomes flat. Then the

ITCZ develops over a kind of SST plateau with no well-defined SST gradient. The reduced SST

gradient locally reduces the wind stress along the equator and the equatorial divergence, which

contributes to damp the development of the equatorial upwelling. Also large differences are found

between the two simulations in the representation of the cloud cover (Figure 10). In TI the

maximum cloud cover follows the ITCZ (it is thus maximum where deep convection occurs).

Figure 10 also shows that during the peak development of the equatorial upwelling these differences

in cloud cover between the two simulations lead to a 80W/m2 difference in net surface shortwave

flux across the equator. Therefore in TI the seasonal evolution of the cloud cover damps the SST

gradient across the equator, and thereby reduces the strength of the wind along the equator and the

strength of the equatorial upwelling.

Figure 11 summarizes the differences along 95°W during the peak development of the

upwelling (August-November). Even though KE exhibits systematic differences with observations,

it exhibits sharp SST and 2 m air temperature gradient across 2°N. Precipitation over the ocean

properly follows the observations. Low level specific humidity increases northwards, with

maximum values around 8°N and minimum values in the southern hemisphere. Several

characteristics found in KE are in good agreement with the data reported by Pyatt et al (2005) along

95°W. Specific humidity is however slightly overestimated (by about 2g/kg) compared to

observation. On the other hand neither the flat SST distribution across the equator nor the uniform

distribution of humidity found in TI are realistic. In this simulation, the convergence of humidity,

5/29/2006 16

the maximum precipitation, and surface cooling by radiative fluxes occur at the same location

(about 4°N).

Seasonal march of the ITCZ in the Atlantic.

The equatorial Atlantic shares some similarities with the eastern Pacific both at the annual and

interannual time scales (Chiang and Vimont 2004). Since in KE the equatorial upwelling develops

in the middle of the basin rather than along the coast (Figure 8), we adjusted a box so as to capture

its seasonal evolution. Figure 9 and 10 consider thus the zonal average of surface conditions

between 30°W and 7°E. The northern limit is provided by the position of the continent in the

northern part of the Golf of Guinea (5°N), so that we limit the latitudinal extension from 12°S to

2°N. Therefore the intensity of the equatorial upwelling is slightly reduced in the observation since

the core of the upwelling is located further east. However, the seasonal evolution of SST and

precipitation is well depicted on these figures. From January to June, warm waters and precipitation

associated with deep convection within the ITCZ expand down to 8°S. From July to November,

cold waters develop from 8°S to 1°N and warm waters and precipitation are located further north

(Figure 9).

For this basin, differences in the representation of the equatorial upwelling and in the location

of the ITCZ between TI and KE are quite similar to those found in the eastern Pacific (Figures 9 and

10). They are even larger. In KE, the seasonal march of the ITCZ over this region follows quite

closely that of the SST. The timing of the north-south migration of the ITCZ across the equator is

well represented. It corresponds to the intensification of the trade winds across the equator and to

the eastward convergence of the wind onto the African continent (Figure 4, right) where monsoon

precipitation occurs (Figure 5). The monsoon flow and the precipitation do not develop as far north

as in the observation, which explains why the zero line of the SST field in figure 9 (right) does not

cross 2°N. During the peak of the upwelling, when warm waters and the ITCZ are in their northern

position, cloud cover intensifies both in the ITCZ and over the cold waters just south of the equator.

5/29/2006 17

In the latter regions these clouds reduce the incoming solar radiation at the surface and contribute to

cool the surface. In TI, the upwelling does not develop and convection is maximum over a flat SST

gradient, like in the Pacific ocean. The overall feature is consistent with the windstress field (Figure

4). In TI, the windstress is not strong enough across and along the equator, and so is the return flow

towards Africa. The convection is more active over the ocean in TI than in KE, as shown by the

500 hPa large-scale vertical velocity in figure 10. The differences in cloud cover also reveal that the

cloud cover develops within the ITCZ in TI. The net effect of these clouds is to damp the

meridional SST gradient in this simulation (Figure 10).

Convection and precipitation in KE are strongly linked to SST. To investigate the link between

SST and rainfall, we compute for each month the correlation between SST and rainfall over the

tropical (30°S-30°N) Atlantic and Pacific for the two simulations (Figure 12). We also compute

this correlation from CMAP (Xie and Arkin 1996) climatology of precipitation and Reynolds’

(1988) climatology of SST. Figure 12 shows that the seasonal evolution is quite satisfactorily

represented in KE. The slight overestimate suggests that this parameterisation is too sensitive to the

local moist static energy which is mainly controlled by the SST (Biasutti et al. 2006). This common

model bias is associated with a tendency for a too strong coupling between rainfall and low level

humidity and a too small coupling with the free troposphere humidity. On the other hand, the

correlation is poor for TI, and even not significant during boreal summer, certainly because SST has

no well defined structure in the equatorial regions in this simulation. During boreal winter the

convergence occurs where SST is maximum south of the equator near South America, which

explains the better correlation found for this period.

5. Interaction between convection, clouds and the large scale circulation

The analyses of the seasonal evolution of the equatorial upwelling in the eastern Pacific and

equatorial Atlantic show that KE and TI exhibit similar differences in the two basins. The seasonal

cycle of the equatorial upwelling is better reproduced in the simulation where the cross equatorial

wind stress is stronger, and in which the wind convergence toward the continent is sustained. This is

5/29/2006 18

in agreement with Xie and Saito (2001), who show that the presence of a bulge of landmass was

important to determine the meridional structure of the ocean ITCZ in the east, and with Fu and

Wang (2001), who identified the meridional wind stress as a critical parameter in the development

of the equatorial upwelling in the east Pacific.

The differences between the two simulations presented here involve the connection between

cross-equatorial trade wind, the monsoonal returned flow over southern North America or West

Africa, and the cloud cover. The way the convection and cloud schemes interact with the large-scale

circulation is to produce such differences is further discussed from atmosphere-alone results

(Hourdin et al. 2006). Then we discuss the ocean feedback.

Link with the characteristics of convection and cloud radiative forcing.

We first compare the behaviour of tropical clouds and convection simulated in the atmospheric

and coupled versions of the model. To segregate regimes of deep convection and upper-level cloud

tops from regimes of shallow convection and low-level clouds, we use the compositing

methodology of Bony et al. (2004), which considers the monthly mean 500 hPa vertical velocity

(ω500) as a proxy for large scale vertical motions in the atmosphere. For the different circulation

regimes, we compare the mean cloud radiative forcing (CRF) and probability distribution function

(PDF) of ω500 derived from coupled and uncoupled experiments with each version of the

convective cloud schemes (Figure 13).

For the different regimes, both the PDF and the CRF of the coupled simulations are similar to

the ones found in the atmosphere alone simulations. In KE the PDF and the CRF are overestimated

(up to 12 Wm-2 for the latter) in regimes of moderate convection. In regimes of precipitating

convection (ω500 < 20 hPa/day), the net CRF at the top of the atmosphere from TI is more negative

(by about 10Wm-2) than that from KE and satellite data. Figure 13c shows that this difference at the

top of the atmosphere partly translates to the surface (the SW CRF at the surface is more negative in

TI than in KE). This suggests that in TI, the cloud cover of the ITCZ contributes more to cool the

surface and to reduce the SST gradient with surrounding areas than in KE. This is consistent with

5/29/2006 19

the colder conditions found in the west Pacific in TI (Figure 1). In subsidence regimes, on the

contrary, TI underestimates the cooling effect of clouds, which also contributes to reduce the SST

gradient. The way convection affects the warming and moistening of the atmospheric column is also

very similar between the coupled and uncoupled simulations, both over the ocean and over land (not

shown).

Analyses of the atmosphere-alone simulations by Hourdin et al. (2006) suggest that the better

representation of land-sea contrasts in rainfall and cloud radiative forcing with the Emanuel’s

scheme comes from the representation of downdraughts over the ocean. In KE convection is

inhibited by downdraughts both over land and over the ocean. In TI the inhibition by downdraughts

is not present over the ocean and convection is not limited, which explains that stronger ascents may

be found in this simulation where convection occurs over ocean (Figure 10). Because of these

differences, convection in TI is better sustained over ocean than over land while in KE it is

comparable over ocean and land, which contributes to reinforce convection over the more

continental northern hemisphere on the average. Figure 14 shows that, as for the atmosphere-alone

simulations, convection is more active in TI over ocean, whereas similar ascending motions are

found in KE over ocean and land. The situation is event more contrasted for TI in the Atlantic. This

effect is more pronounced in the coupled simulations because of the feedback with SST, as

explainsed below.

Role of the monsoon flow and low level clouds in the eastern part of the Pacific and Atlantic

In KE, several feedbacks from the atmospheric circulation come into play in the coupled model

to cool down SST along the equator and enhance the SST gradient in KE in the East Pacific and

Atlantic. In TI the feedbacks act to damp the SST gradients and the ITCZ remains in a rather south

position. These feedbacks involve the ocean circulation and the cloud cover.

Figure 15 shows that in August-September, during the peak development of the upwelling, the

zonal currents in the first layer of the ocean model (first 10 m) are smaller in KE than in TI in the

5/29/2006 20

eastern Pacific. The reversal of the westward south equatorial current and the eastward north

equatorial current is located at 4°N against 3°N in TI. On the contrary, the meridional component of

the current, and the vertical velocity at the base of the first layer are stronger in KE across the

equator (Figure 15). Upwelled waters cool the surface, and the meridional component of the current

advects northwards cold water from the south of the equator. The meridional windstress component

across the equator and its northwards convergence are thus two critical features that help maintain

cold water along the equator and a sharp SST gradient to the north of it. This SST gradient locally

enhances the windstress. The role of vertical and meridional temperature advection in cooling the

surface layer is less active in TI. Also since the thermocline is more diffuse in TI upwelled waters

are warmer, reducing the cooling effect of the upwelling. Note that in the Atlantic ocean, the

components of the currents are more similar between the two simulations. The major differences

come from warmer upwelled waters in the first layer.

As seen in section 4, the cloud cover further contributes to enhance the SST gradient in KE and

to damp it in TI. In KE, the SST gradient between the cold equatorial waters and the regions further

north is reinforced by the presence of low level clouds over cold waters. The stronger SST gradient

maintains the wind stress and the development of the equatorial upwelling. In TI, the development

of the equatorial upwelling is not sufficient to maintain a sharp cross equatorial SST gradient that

would reinforce the windstress. In addition convective clouds strongly cool down the surface where

convection occurs, which further contributes to weaken the SST gradient, and to reinforce the

southward shift of the ITCZ.

6. Conclusion

The comparison of two coupled ocean-atmosphere simulations performed with different

convective clouds schemes allows us to investigate how the representation of clouds and convection

interacts with the large-scale tropical atmospheric and oceanic circulations.

We show that, to first order, the major characteristics of surface climate simulated by the

coupled model are primarily controlled by atmospheric processes model. Indeed the distribution of

5/29/2006 21

cloud radiative forcing and convection regimes in the tropical regions is the same between the

coupled simulations and simulations with the atmospheric model forced by prescribed SST. In

addition several features of the atmospheric simulations are only slightly affected by the coupling

with the ocean. This includes in particular the annual mean and the seasonal evolution of the

atmospheric temperature, of the precipitation, of the cloud radiative forcing and of the surface wind.

The key role of atmospheric processes on coupled simulations has been noted in other studies

considering several ocean models and various aspects of the simulated climate. From a set of

several state of the art coupled model simulations where models shared either the same atmosphere

or the same ocean, Guilyardi et al. (2004) found that the atmosphere has a dominant role in setting

El Niño characteristics (periodicity and base amplitude) and errors (regularity). From an analysis of

climate sensitivity under global warming with different combinations of the NCAR models, Meehl

et al. (2004) concluded the atmospheric model "manages'' the relevant global feedbacks including

sea ice albedo, water vapor, and clouds, and that for global sensitivity measures, the ocean, sea ice,

and land surface play secondary roles, even though differences in these components can be

important for regional climate changes.

In the simulations presented here, differences in the atmospheric simulations propagate down to

about 1000m in the ocean interior. The simulated mean state and the seasonal cycle of the SST in

most part of the equatorial ocean differ strongly depending on the convection and clouds scheme

used. Our results show that the cross equatorial wind in the eastern Pacific and in the Atlantic

during boreal summer is critical to maintain the equatorial upwelling, and the advection of cold

waters to the north of the equator. The simulation that best represent the position of the ITCZ in

these region is the simulation for which this cross equatorial circulation is more active, which

maintains a sharp SST gradient in the northern hemisphere. This simulation also exhibits a weaker

zonal circulation along the equator. The dominant feature is the link between large scale ocean and

continental patterns. In particular, these patterns are connected during summer through the large

scale monsoon flow in the eastern equatorial Pacific and Atlantic oceans. Therefore a good

5/29/2006 22

representation of the land sea contrast and of the large scale circulation in the tropics is a key to

properly simulate the seasonal cycle over the eastern part of the ocean.

Our analysis also shows that the interaction between convection, clouds and the large scale

tropical circulation dominate the differences at the seasonal time scale. In the simulation where

convection heating and moistening is inhibited by downdrafts both over the continent and the ocean

(KE), the monsoon flow converging in south North America and Africa is better reproduced, and

this reinforces the maintenance of the inland moisture convergence. In the simulation (TI) where

convection is not inhibited by downdraft over the ocean, convection and moisture convergence are

too strong over the ocean, and favor the large scale convergence over the ocean at the expanse of

the continent. The ITCZ is thus located further south all year round in this simulation. This

mechanism is reinforced by the radiative impact of the cloud cover. In one case (TI) the cloud cover

tends to cool the surface in (warm) regions of convection whereas in the other case (KE), low-level

clouds in (cold) subsidence regions reinforce the SST gradient between convective regions and their

surroundings, which reinforces the windstress and the equatorial upwelling. In both cases the

feedback from the ocean circulation leads to SST patterns that enhance these basic mechanisms.

These results clearly show that understanding the coupled system requires to consider in an

integrated way the evolution of the seasonal cycle over land and over ocean, since they are

interconnected through the atmospheric large-scale circulation. The investigation of the relative

strength of deep convection between ocean and land in several models is thus likely to help us to

understand why different climate models produce different mean climate states and seasonal cycles

in the tropics. Our analysis also suggests that a model assessment should be based on the evaluation

both on the assessment of the model's climatology and on the assessment of on both the comparison

of simulated climate fields with observations, as well as on key mechanisms. The comparison with

climatology reveals that both simulations have serious biases compared to observations. However,

when going into the details, it becomes clear that the KE simulation better represents several major

features of the mean climate and of the seasonal cycle in the tropical regions. The differences in the

parameterisation between these two simulations also have an impact on the interannual variability

5/29/2006 23

simulated by the coupled model. This will deserve future investigation, and will be considered in a

forthcoming study.

Acknowledgement. We would like to thank all IPSL people who participate to the development of

the IPSL_CM4 model. We also thank Laurent Fairhead for the introduction of the convection

schemes in the LMDZ model, and Ionela Musat for the adjustments of the climatology. Computer

time was provided by Centre National de la Recherche Scientifique (IDRIS computing center) and

Commissariat à l’Energie Atomique (centre CCRT computing center).

5/29/2006 24

References Barkstrom, B. R. (1984). The earth radiation budget experiment (erbe). Bull. Am. Meteorol. Soc 65:

1170-1185.

Biasutti, M., A. Sobel and Y. Kushnir (2006). Agcm precipitation biases in the tropical atlantic.

Journal of Climate 19(No. 6): 935-958.

Bony, S., J. Dufresne, H. Le Treut, J. Morcrette and C. Senior (2004). On dynamic and

thermodynamic components of cloud changes. CLIMATE DYNAMICS 22(2-3): 71-86.

Bony, S. and K. A. Emanuel (2001). A parameterization of the cloudiness associated with cumulus

convection; evaluation using toga coare data. Journal of the Atmospheric Sciences 58(21):

3158-3183.

Braconnot, P. (1998). Tests de sensibilité avec le modèle d'atmosphère du lmd, en vue d'améliorer

le couplage avec l'océan. note technique IPSL 2: 39.

Braconnot, P., J. Joussaume, O. Marti and N. de Noblet (2000). Impact of ocean and vegetation

feedback on 6 ka monsoon changes. Canada, WCRP-111, WMO/TD-No. 1007.

Chiang, J. and D. Vimont (2004). Analogous pacific and atlantic meridional modes of tropical

atmosphere-ocean variability. JOURNAL OF CLIMATE 17(21): 4143-4158.

Davey, M. K., M. Huddleston, K. R. Sperber, P. Braconnot, F. Bryan, D. Chen, R. A. Colman, C.

Cooper, U. Cubasch, P. Delecluse, D. DeWitt, L. Fairhead, G. Flato, C. Gordon, T. Hogan, M.

Ji, M. Kimoto, A. Kitoh, T. R. Knutson, M. Latif, H. Le Treut, T. Li, S. Manabe, C. R.

Mechoso, G. A. Meehl, S. B. Power, E. Roeckner, L. Terray, A. Vintzileos, R. Voss, B. Wang,

W. M. Washington, I. Yoshikawa, J. Y. Yu, S. Yukimoto and S. E. Zebiak (2002). Stoic: A

study of coupled model climatology and variability in tropical ocean regions. Climate

Dynamics 18(5): 403-420.

Dufresne, J.-L. and J.-Y. Grandpeix (1996). Raccordement des modèles thermodynamiques de

glace, d'océan et d'atmosphère. Note Interne 205, L.M.D.

5/29/2006 25

Dufresne, J., J. Quaas, O. Boucher, S. Denvil and L. Fairhead (2005). Contrasts in the effects on

climate of anthropogenic sulfate aerosols between the 20th and the 21st century.

GEOPHYSICAL RESEARCH LETTERS 32(21): -.

Dufresne, J. L., P. Friedlingstein, M. Berthelot, L. Bopp, P. Ciais, L. Fairhead, H. Le Treut and P.

Monfray (2002). On the magnitude of positive feedback between future climate change and the

carbon cycle. Geophysical Research Letters 29(10): art. no.-1405.

Emanuel, K. A. (1993). A scheme for representing cumulus convection in large-scale models.

Journal of Atmospheric Science 48(2313-2335).

Fichefet, T. and M. A. M. Maqueda (1997). Sensitivity of a global sea ice model to the treatment of

ice thermodynamics and dynamics. Journal of Geophysical Research-Oceans 102(C6): 12609-

12646.

Fu, X. and B. Wang (2001). A coupled modeling study of the seasonal cycle of the pacific cold

tongue. Part i: Simulation and sensitivity experiments. JOURNAL OF CLIMATE 14(5): 765-

779.

Gordon, C., A. Rosati and R. Gudgel (2000). Tropical sensitivity of a coupled model to specified

isccp low clouds. JOURNAL OF CLIMATE 13(13): 2239-2260.

Guilyardi, E., S. Gualdi, J. Slingo, A. Navarra, P. Delecluse, J. Cole, G. Madec, M. Roberts, M.

Latif and L. Terray (2004). Representing el nino in coupled ocean-atmosphere gcms: The

dominant role of the atmospheric component. Journal of Climate 17(24): 4623-4629.

Hourdin, F., I. Musat, S. Bony, P. Braconnot, F. Codron, J. L. Dufresne, L. Fairhead, M. A.

Filiberti, P. Friedlingstein, J. Y. Grandpeix, G. Krinner, P. Le Van, Z. X. Li and F. Lott (2006).

The lmdz4 general circulation model: Climate performance and sensitivity to parametrized

physics with emphasis on tropical convection. Climate Dynamics: in press.

IPCC (2001). Climate change 2001, the scientific basis, Cambridge University press.

Krinner, G., N. Viovy, N. de Noblet-Ducoudre, J. Ogee, J. Polcher, P. Friedlingstein, P. Ciais, S.

Sitch and I. Prentice (2005). A dynamic global vegetation model for studies of the coupled

atmosphere-biosphere system. GLOBAL BIOGEOCHEMICAL CYCLES 19(1): -.

5/29/2006 26

Latif, M., K. Sperber, J. Arblaster, P. Braconnot, D. Chen, A. Colman, U. Cubasch, C. Cooper, P.

Delecluse, D. DeWitt, L. Fairhead, G. Flato, T. Hogan, M. Ji, M. Kimoto, A. Kitoh, T. Knutson,

H. Le Treut, T. Li, S. Manabe, O. Marti, C. Mechoso, G. Meehl, S. Power, E. Roeckner, J.

Sirven, L. Terray, A. Vintzileos, R. Voss, B. Wang, W. Washington, I. Yoshikawa, J. Yu and S.

Zebiak (2001). Ensip: The el nino simulation intercomparison project. Climate Dynamics 18(3-

4): 255-276.

Levitus, S. (1982). Climatological atlas of the world ocean, NOAA Professional paper, Washington

D.C.

Li, X. (1999). Ensemble atmospheric gcm simulations of climate interannual variability from 1979

to 1994. Journal of Climate 12: 986-1001.

Li, Z. X. and S. Conil (2003). A 1000-year simulation with the ipsl ocean-atmosphere coupled

model. Annals of Geophysics 46(1): 39-46.

Ma, C. C., C. R. Mechoso, A. W. Robertson and A. Arakawa (1996). Peruvian stratus clouds and

the tropical pacific circulation: A coupled ocean-atmosphere gcm study. Journal of Climate

9(7): 1635-1645.

Madec, G., P. Delecluse, M. Imbart and C. Levy (1998). Opa 8.1 ocean general circulation model

reference manual. Note du Pôle de modélisation, Institut Pierre-Simon Laplace 11: 94 pp.

Marti, O., P. Braconnot, J. Bellier, R. Benshila, S. Bony, P. Brockmann, P. Cadule, A. Caubel, S.

Denvil, J. L. Dufresne, L. Fairhead, M. A. Filiberti, M.-A. Foujols, T. Fichefet, P.

Friedlingstein, H. Goosse, J. Y. Grandpeix, F. Hourdin, G. Krinner, C. Lévy, G. Madec, I.

Musat, N. deNoblet, J. Polcher and C. Talandier (2005). The new ipsl climate system model:

Ipsl-cm4. Note du Pôle de Modélisation n 26, ISSN 1288-1619.

Meehl, G. A., W. M. Washington, J. M. Arblaster and A. X. Hu (2004). Factors affecting climate

sensitivity in global coupled models. Journal of Climate 17(7): 1584-1596.

Nigam, S. and Y. Chao (1996). Evolution dynamics of tropical ocean-atmosphere annual cycle

variability. Journal of Climate 9(12): 3187-3205.

5/29/2006 27

Pyatt, H. E., B. A. Albrecht, C. Fairall, J. E. Hare, N. Bond, P. Minnis and J. K. Ayers (2005).

Evolution of marine atmospheric boundary layer structure across the cold tongue-itcz complex.

Journal of Climate 18(5): 737-753.

Reynolds, R. W. (1988). A real time global sea-surface temperature analysis. Journal of Climate 1:

75-86.

Swingedouw, D., P. Braconnot, P. Delecluse, E. Guilyardi and O. Marti (2005). Sensitivity of the

atlantic thermohaline circulation to global freshwater forcing. Climate Dynamics.

Swingedouw, D., P. Braconnot and O. Marti (2006). Sensitivity of the atlantic meridional

overturning circulation to the melting from northern glaciers in climate change experiments.

Geophys. Res. Lett. 33, No 7.

Terray, L., E. Sevault, E. Guilyardi and O. Thual (1995). The oasis coupler user guide version 2.0.

Cerfacs technical report TR/CMGC: 95-46.

Tiedke, M. (1989). A comprehensive mass flux scheme for cumulus parameterization in large-scale

models. Monthly Weather Review 117: 1179-1800.

Trenberth, K. and A. Salomon (1994). The global heat balance: Heat transports in the atmosphere

and ocean. Climate Dynamics 10: 107-134.

Wang, B. (1994). On the annual cycle in the tropical eastern central pacific. Journal of Climate

7(12): 1926-1942.

Xie, P. and P. Arkin (1996). Analyses of global monthly precipitation using gauge observations,

satellite estimates, and numerical model predictions. JOURNAL OF CLIMATE 9(4): 840-858.

Xie, S. and K. Saito (2001). Formation and variability of a northerly itcz in a hybrid coupled agcm:

Continental forcing and oceanic-atmospheric feedback. JOURNAL OF CLIMATE 14(6): 1262-

1276.

Yu, J. Y. and C. R. Mechoso (1999). Links between annual variations of peruvian stratocumulus

clouds and of sst in the eastern equatorial pacific. Journal of Climate 12(11): 3305-3318.

5/29/2006 28

Figure caption.

Figure 1: Annual mean sea surface temperature difference (°C) between a) KE and the Levitus

(Levitus 1982) climatology and b) Ti and KE. Values larger (lower) than 2°C (-2°C) are dark (light)

shaded. Dashed (solid) lines stand for negative (positive) values.

Figure 2: Annual mean atmosphere and ocean temperature (°C), zonally averaged and plotted as a

function of latitude and depth, a) difference between KE and climatology, b) difference between Ti

and KE. Note that the first 1000 m of the ocean are expended. Values larger (lower) than 2°C (-2°C)

are dark (light) shaded.

Figure 3: Equatorial section of the first 350 meters of the ocean averaged between 2°N and 2°S for

a) Levitus’ climatology, b) KE and c) TI. Isolines are plotted every 2°C. Values higher than 14°C

are shaded.

Figure 4: January (left) and July (right) wind stress (in Pa) between 30°N and 30°S for a) ERS data,

b) KE, and c) Ti.

Figure 5: January (left) and July (right) precipitation (mm/d) for a) the CMAP climatology (Xie and

Arkin 1996), b) KE, c) Ti, d) differences between KE and climatology and e) differences between

Ti and KE. In all plots isolines are plotted every 2mm/d (the zero line is omitted).

Figure 6: Comparison of annual mean evolution of SST (°C) and precipitation (mmd-1) for KE

(solid line), TI (dashed line) and observations (dotted line), for several boxes over the tropical

ocean. The box limits are identified as dotted rectangles on the map. The two boxes represented by

solid lines correspond to the limits of the regions used to in section 4.

Figure 7: Comparison of the characteristics of the simulated mean seasonal cycle of SST with the

Reynold’s (1988). a) Annual mean cycle as estimated from data, b) and c) ratio of the difference

between the simulated and observed annual mean SST amplitude (defined as the maximum monthly

temperature minus the minimum monthly temperature) to the observed amplitude for TI and KE

respectively. d) and e) correlation between the simulated seasonal cycle at each model grid point

with the observation for TI and KE. The data have been interpolated on the model grid. Solid lines

stand for positive values, dashed lines for negative values and, in c) and b), the heavy solid line

represent the 0 correlation.

5/29/2006 29

Figure 8: Seasonal mean evolution (annual mean removed) of SST (°C) along the equator averaged

between 2°N-2°S for a) Reynolds ‘ (1988) climatology b) KE and c) TI.

Figure 9: Annual mean evolution of SST (annual mean of the box removed; isolines; °C) and

precipitation (shading; mm/d) averaged from a)-c) 90°W to 110°W in the Pacific and e)-f) 30°W to

7°E in the Atlantic and plotted as a function of latitude and time for a) the Reynold’s SST and

CMAP precipitation climatologies, b) KE and c) TI.

Figure 10: Same as figure 9 but for a)-c) and e)-g) the seasonal mean evolution of ω500hPa (hPa d-

1, isolines) and total cloud fraction (shading) and d) and h) the difference between the surface net

shortwave flux between TI and KE. For shortwave, the annual mean value over the region was

subtracted from each simulation to better emphasise the gradients

Figure 11: Cross equatorial section averaged from 90°W to 110°W and from August to November

of a) SST (°C) b) 2 m air temperature (t2m,°C) c) precipitation (Precip, mm/day), d) 2m air specific

humidity (q2m, g/kg), e) net radiation at the surface (Wm-2, positive downward), and f) the

meridional latent heat transport integrated over the atmospheric column (vq, gm-1s-1). In each graph,

the solid line represents KE and the dashed line TI. In a) and c) the dotted line stands for the

observations (CMAP for precipitation and Reynolds for SST).

Figure 12: spatial correlation between precipitation and SST computed for each month between

30°N and 30°S for a) the Pacific ocean and b) the Atlantic ocean. The solide lines represent the

estimation from KE, the dashed line the results from TI and the dotted line an estimate from the

Reynold’s (1988) and CMAP climatologies.

Figure 13: Regime sorted plots as a function of ω500 in hPa/d between 30°N and 30°S over the

ocean for a) the probability distribution function of ω500, b) the net cloud radiative forcing a the top

of the atmosphere (Wm-2), c) the net cloud radiative forcing at the ocean surface (Wm-2). In all

graphs the black lines stand for simulations with the Tiedke convection scheme and the red lines for

simulations with the Emanuel convection scheme. In figure a) and b) the coupled simulations are

represented by the solid lines and AMIP type simulations with the atmospheric model by the dotted

lines. For comparison two estimates from ERA40 and NCEP reanalyses for ω500 and ERBE

(Barkstrom 1984) for cloud radiative forcing are included.

Figure 14: Bar charts representing the 500 hPa (hPas-1) vertical velocity averaged over land and

ocean from June to October in TI and KE for the eastern Pacific and Atlantic. The boxes are

respectively 100°W-80°W; 0°N-20°N for the Pacific and 20°W-7°E; 0°N-20°N for the Atlantic.

5/29/2006 30

Figure 15: Ocean currents for the first upper layer of the ocean model and SST in August in the

Pacific and Atlantic. a) zonal current in the first model layer (m.s-1), b) meridional current in the

first model layer (m.s-1), c) vertical velocity at the base of the first layer (m.d-1), and d) SST (°C). In

a), b) c), dark grey represents positive values and light grey negative values.

Figure 1: Annual mean sea surface temperature difference (°C) between a) KE and the Levitus

(Levitus, 1982) climatology and b) Ti and KE. Values larger (lower) than 2°C (-2°C) are dark (light)

shaded. Dashed (solid) lines stand for negative (positive) values.

Figure 2: Annual mean atmosphere and ocean temperature (°C), zonally averaged and plotted as a

function of latitude and depth, a) difference between KE and climatology, b) difference between Ti

and KE. Note that the first 1000 m of the ocean are expended. Values larger (lower) than 2°C (-2°C)

are dark (light) shaded.

Figure 3:

Figure 3: Equatorial section of the first 350 meters of the ocean averaged between 2°N and 2°S for a)

Levitus’ climatology, b) KE and c) TI. Isolines are plotted every 2°C. Values higher than 14°C are

shaded.

Figure 4: January (left) and July (right) wind stress (in Pa) between 30°N and 30°S for a) ERS data, b) KE, and c) Ti.

Figure 5: January (left) and July (right) precipitation (mm/d) for a) the CMAP climatology (Xie and

Arkin 1996), b) KE, c) Ti, d) differences between KE and climatology and e) differences between Ti

and KE. In all plots isolines are plotted every 2mm/d (the zero line is omitted).

Figure 6: Comparison of annual mean evolution of SST (°C) and precipitation (mm d-1) for KE (solid

line), TI (dashed line) and observations (dotted line), for several boxes over the tropical ocean. The

box limits are identified as dotted rectangles on the map. The two boxes represented by solid lines

correspond to the limits of the regions used to in section 4.

Figure 7: Comparison of the characteristics of the simulated mean seasonal cycle of SST with the

Reynold’s (1988). a) Annual mean cycle as estimated from data, b) and c) ratio of the difference

between the simulated and observed annual mean SST amplitude (defined as the maximum monthly

temperature minus the minimum monthly temperature) to the observed amplitude for TI and KE

respectively. d) and e) correlation between the simulated seasonal cycle at each model grid point with

the observation for TI and KE. The data have been interpolated on the model grid. Solid lines stand

for positive values, dashed lines for negative values and, in c) and b), the heavy solid line represent the

0 correlation.

Figure 8: Seasonal mean evolution (annual mean removed) of SST (°C) along the equator averaged

between 2°N-2°S for a) Reynolds ‘(1988) climatology b) KE and c) TI.

Figure 9: Annual mean evolution of SST (annual mean of the box removed; isolines; °C) and

precipitation (shading; mm/d) averaged from a)-c) 90°W to 110°W in the Pacific and e)-f) 30°W to

7°E in the Atlantic and plotted as a function of latitude and time for a) the Reynold’s SST and CMAP

precipitation climatologies, b) KE and c) TI.

Figure 10 : Same as figure 9 but for a)-c) and e)-g) the seasonal mean evolution of ω500hPa (hPad-1,

isolines) and total cloud fraction (shading) and d) and h) the difference between the surface net

shortwave flux between TI and KE. For shortwave, the annual mean value over the region was

subtracted from each simulation to better emphasise the gradients

Figure 11: Cross equatorial section averaged from 90°W to 110°W and from August to November of

a) SST (°C) b) 2 m air temperature (t2m,°C) c) precipitation (Precip, mm/day), d) 2m air specific

humidity (q2m, g/kg), e) net radiation at the surface (Wm-2, positive downward), and f) the meridional

latent heat transport integrated over the atmospheric column (vq, gm-1s-1). In each graph, the solid line

represents KE and the dashed line TI. In a) and c) the dotted line stands for the observations (CMAP

for precipitation and Reynolds for SST).

Figure 12: spatial correlation between precipitation and SST computed for each month between 30°N

and 30°S for a) the Pacific ocean and b) the Atlantic ocean. The solide lines represent the estimation

from KE, the dashed line the results from TI and the dotted line an estimate from the Reynold’s (1988)

and CMAP climatologies.

Figure 13: Regime sorted plots as a function of ω500 in hPa/d between 30°N and 30°S over the ocean

for a) the probability distribution function of ω500, b) the net cloud radiative forcing a the top of the

atmosphere (Wm-2), c) the net cloud radiative forcing at the ocean surface (Wm-2). In all graphs the

black lines stand for simulations with the Tiedke convection scheme and the red lines for simulations

with the Emanuel convection scheme. In figure a) and b) the coupled simulation are represented by the

solid lines and AMIP type simulation with the atmospheric model by the dotted lines. For comparison

two estimates from ERA40 and NCEP reanalyses for ω500 and ERBE (Barkstrom, 1984) for cloud

radiative forcing are included.

vertical velocity at 500 hPA: -100°W-80°W Eq-20°N

-0,02

-0,015

-0,01

-0,005

0land ocean

KETI

Vertical velocity at 500 hPa : 20°W-7°E Eq-20°N

-0,06

-0,05

-0,04

-0,03

-0,02

-0,01

0land ocean

KETI

Figure 14: Bar charts representing the 500 hPa (hPas-1) vertical velocity averaged over land and ocean

from June to October in TI and KE for the eastern Pacific and Atlantic. The boxes are respectively

100°W-80°W; 0°N-20°N for the Pacific and 20°W-7°E; 0°N-20°N for the Atlantic.

Figure 15: Ocean currents for the first upper layer of the ocean model and SST in August in the Pacific

and Atlantic. a) zonal current in the first model layer (m.s-1), b) meridional current in the first model

layer (m.s-1), c) vertical velocity at the base of the first layer (m.d-1), and d) SST (°C). In a), b) c), dark

grey represents positive values and light grey negative values.